[Bug libc/15794] New: move AT_* constants definition from elf.h into sysdeps (bits) directory

[Bug libc/15794] New: move AT_* constants definition from elf.h into sysdeps (bits) directory

[Petr.Salinger at seznam dot cz]

http://sourceware.org/bugzilla/show_bug.cgi?id=15794

            Bug ID: 15794
           Summary: move AT_* constants definition from elf.h into sysdeps
                    (bits) directory
           Product: glibc
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: libc
          Assignee: unassigned at sourceware dot org
          Reporter: Petr.Salinger at seznam dot cz
                CC: drepper.fsp at gmail dot com

Created attachment 7124
  --> http://sourceware.org/bugzilla/attachment.cgi?id=7124&action=edit
proposed patch

Currently the installed <elf.h> contains many Linux specific AT_* constants.
Especially with introduction of getauxval() it would be handy to allow
overriding them.

Please move this into separate bits header <bits/auxv.h>.

At least the FreeBSD kernel uses a different set.
See http://svnweb.freebsd.org/base/head/sys/x86/include/elf.h

And I doubt hurd also provides all of them.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/15794] move AT_* constants definition from elf.h into sysdeps (bits) directory

[neleai at seznam dot cz]

http://sourceware.org/bugzilla/show_bug.cgi?id=15794

--- Comment #1 from Ondrej Bilka <neleai at seznam dot cz> ---
> 
> Currently the installed <elf.h> contains many Linux specific AT_* constants.
> Especially with introduction of getauxval() it would be handy to allow
> overriding them.
> 
> Please move this into separate bits header <bits/auxv.h>.
> 
> At least the FreeBSD kernel uses a different set.
> See http://svnweb.freebsd.org/base/head/sys/x86/include/elf.h
> 
> And I doubt hurd also provides all of them.
> 
As this is patch please send it to libc-alpha@sourceware.org to get
comments there.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

Re: [Bug libc/15794] New: move AT_* constants definition from elf.h into sysdeps (bits) directory

[Ondřej Bílka]

> 
> Currently the installed <elf.h> contains many Linux specific AT_* constants.
> Especially with introduction of getauxval() it would be handy to allow
> overriding them.
> 
> Please move this into separate bits header <bits/auxv.h>.
> 
> At least the FreeBSD kernel uses a different set.
> See http://svnweb.freebsd.org/base/head/sys/x86/include/elf.h
> 
> And I doubt hurd also provides all of them.
> 
As this is patch please send it to libc-alpha@sourceware.org to get
comments there. 

[Bug libc/15794] move AT_* constants definition from elf.h into sysdeps (bits) directory

[Petr.Salinger at seznam dot cz]

http://sourceware.org/bugzilla/show_bug.cgi?id=15794

--- Comment #3 from Petr.Salinger at seznam dot cz ---
I would rather consider the AT_* constants as constants in errno.h
There is a common base, with the same numbers (0-9).
But there are also some constants with same meaning and different numbers
and constants with same number but completely different meaning.

The original problem is
http://bugs.debian.org//cgi-bin/bugreport.cgi?bug=717912


https://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git/tree/include/uapi/linux/auxvec.h

#define AT_NULL   0    /* end of vector */
#define AT_IGNORE 1    /* entry should be ignored */
#define AT_EXECFD 2    /* file descriptor of program */
#define AT_PHDR   3    /* program headers for program */
#define AT_PHENT  4    /* size of program header entry */
#define AT_PHNUM  5    /* number of program headers */
#define AT_PAGESZ 6    /* system page size */
#define AT_BASE   7    /* base address of interpreter */
#define AT_FLAGS  8    /* flags */
#define AT_ENTRY  9    /* entry point of program */
#define AT_NOTELF 10    /* program is not ELF */
#define AT_UID    11    /* real uid */
#define AT_EUID   12    /* effective uid */
#define AT_GID    13    /* real gid */
#define AT_EGID   14    /* effective gid */
#define AT_PLATFORM 15  /* string identifying CPU for optimizations */
#define AT_HWCAP  16    /* arch dependent hints at CPU capabilities */
#define AT_CLKTCK 17    /* frequency at which times() increments */
/* AT_* values 18 through 22 are reserved */
#define AT_SECURE 23   /* secure mode boolean */
#define AT_BASE_PLATFORM 24    /* string identifying real platform, may
                 * differ from AT_PLATFORM. */
#define AT_RANDOM 25    /* address of 16 random bytes */
#define AT_HWCAP2 26    /* extension of AT_HWCAP */

#define AT_EXECFN  31    /* filename of program */


http://svnweb.freebsd.org/base/head/sys/x86/include/elf.h:
/* Values for a_type. */
#define    AT_NULL        0    /* Terminates the vector. */
#define    AT_IGNORE    1    /* Ignored entry. */
#define    AT_EXECFD    2    /* File descriptor of program to load. */
#define    AT_PHDR        3    /* Program header of program already loaded. */
#define    AT_PHENT    4    /* Size of each program header entry. */
#define    AT_PHNUM    5    /* Number of program header entries. */
#define    AT_PAGESZ    6    /* Page size in bytes. */
#define    AT_BASE        7    /* Interpreter's base address. */
#define    AT_FLAGS    8    /* Flags (unused for i386). */
#define    AT_ENTRY    9    /* Where interpreter should transfer control. */
#define    AT_NOTELF    10    /* Program is not ELF ?? */
#define    AT_UID        11    /* Real uid. */
#define    AT_EUID        12    /* Effective uid. */
#define    AT_GID        13    /* Real gid. */
#define    AT_EGID        14    /* Effective gid. */
#define    AT_EXECPATH    15    /* Path to the executable. */
#define    AT_CANARY    16    /* Canary for SSP. */
#define    AT_CANARYLEN    17    /* Length of the canary. */
#define    AT_OSRELDATE    18    /* OSRELDATE. */
#define    AT_NCPUS    19    /* Number of CPUs. */
#define    AT_PAGESIZES    20    /* Pagesizes. */
#define    AT_PAGESIZESLEN    21    /* Number of pagesizes. */
#define    AT_TIMEKEEP    22    /* Pointer to timehands. */
#define    AT_STACKPROT    23    /* Initial stack protection. */



http://www.openbsd.org/cgi-bin/cvsweb/src/sys/sys/exec_elf.h
enum AuxID {
    AUX_null = 0,
    AUX_ignore = 1,
    AUX_execfd = 2,
    AUX_phdr = 3,            /* &phdr[0] */
    AUX_phent = 4,            /* sizeof(phdr[0]) */
    AUX_phnum = 5,            /* # phdr entries */
    AUX_pagesz = 6,            /* PAGESIZE */
    AUX_base = 7,            /* ld.so base addr */
    AUX_flags = 8,            /* processor flags */
    AUX_entry = 9,            /* a.out entry */
    AUX_sun_uid = 2000,        /* euid */
    AUX_sun_ruid = 2001,        /* ruid */
    AUX_sun_gid = 2002,        /* egid */
    AUX_sun_rgid = 2003        /* rgid */
};

http://cvsweb.netbsd.org/bsdweb.cgi/~checkout~/src/sys/sys/exec_elf.h
/* a_type */
#define AT_NULL        0    /* Marks end of array */
#define AT_IGNORE    1    /* No meaning, a_un is undefined */
#define AT_EXECFD    2    /* Open file descriptor of object file */
#define AT_PHDR        3    /* &phdr[0] */
#define AT_PHENT    4    /* sizeof(phdr[0]) */
#define AT_PHNUM    5    /* # phdr entries */
#define AT_PAGESZ    6    /* PAGESIZE */
#define AT_BASE        7    /* Interpreter base addr */
#define AT_FLAGS    8    /* Processor flags */
#define AT_ENTRY    9    /* Entry address of executable */
#define AT_DCACHEBSIZE    10    /* Data cache block size */
#define AT_ICACHEBSIZE    11    /* Instruction cache block size */
#define AT_UCACHEBSIZE    12    /* Unified cache block size */
#define AT_STACKBASE    13    /* Base address of the main thread */

    /* Vendor specific */
#define AT_MIPS_NOTELF    10    /* XXX a_val != 0 -> MIPS XCOFF executable */

#define AT_EUID        2000    /* euid (solaris compatible numbers) */
#define AT_RUID        2001    /* ruid (solaris compatible numbers) */
#define AT_EGID        2002    /* egid (solaris compatible numbers) */
#define AT_RGID        2003    /* rgid (solaris compatible numbers) */

    /* Solaris kernel specific */
#define AT_SUN_LDELF    2004    /* dynamic linker's ELF header */
#define AT_SUN_LDSHDR    2005    /* dynamic linker's section header */
#define AT_SUN_LDNAME    2006    /* dynamic linker's name */
#define AT_SUN_LPGSIZE    2007    /* large pagesize */

    /* Other information */
#define AT_SUN_PLATFORM 2008    /* sysinfo(SI_PLATFORM) */
#define AT_SUN_HWCAP    2009    /* process hardware capabilities */
#define AT_SUN_IFLUSH    2010    /* do we need to flush the instruction cache?
*/
#define AT_SUN_CPU    2011    /* CPU name */
    /* ibcs2 emulation band aid */
#define AT_SUN_EMUL_ENTRY 2012    /* coff entry point */
#define AT_SUN_EMUL_EXECFD 2013 /* coff file descriptor */
    /* Executable's fully resolved name */
#define AT_SUN_EXECNAME 2014

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/15794] move AT_* constants definition from elf.h into sysdeps (bits) directory

[joseph at codesourcery dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=15794

--- Comment #4 from joseph at codesourcery dot com <joseph at codesourcery dot com> ---
If a value is considered like errno.h, it shouldn't be included in elf.h, 
directly or indirectly, at all, as elf.h should be equally usable for 
cross tools as for native ones.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/15794] move AT_* constants definition from elf.h into sysdeps (bits) directory

[Petr.Salinger at seznam dot cz]

http://sourceware.org/bugzilla/show_bug.cgi?id=15794

Petr.Salinger at seznam dot cz changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Attachment #7124|0                           |1
        is obsolete|                            |

--- Comment #5 from Petr.Salinger at seznam dot cz ---
Created attachment 7125
  --> http://sourceware.org/bugzilla/attachment.cgi?id=7125&action=edit
v2

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/15794] move AT_* constants definition from elf.h into sysdeps (bits) directory

[Petr.Salinger at seznam dot cz]

http://sourceware.org/bugzilla/show_bug.cgi?id=15794

--- Comment #6 from Petr.Salinger at seznam dot cz ---
In v2 patch AT_* constants are moved from <elf.h> into <bits/auxv.h>,
this new bits header is included from <sys/auxv.h> for userspace usage
and from ldsodefs.h for ld.so build.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/15794] move AT_* constants definition from elf.h into sysdeps (bits) directory

[adconrad at 0c3 dot net]

https://sourceware.org/bugzilla/show_bug.cgi?id=15794

Adam Conrad <adconrad at 0c3 dot net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |adconrad at 0c3 dot net

--- Comment #7 from Adam Conrad <adconrad at 0c3 dot net> ---
The problem with this patch is that any code that previously included elf.h to
get the AT_ defines is now broken (*cough*GCC*cough*).  Are you expecting that
any such code should have an include <sys/auxv.h>, or should elf.h be pulling
in bits/auxv.h for backward compat?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/15794] move AT_* constants definition from elf.h into sysdeps (bits) directory

[adconrad at 0c3 dot net]

https://sourceware.org/bugzilla/show_bug.cgi?id=15794

--- Comment #8 from Adam Conrad <adconrad at 0c3 dot net> ---
Of note, it seems that on (at least) Solaris, the AT_ defines come from
sys/auxv.h, not elf.h, so there's some precedent here for moving them, and
perhaps not much code would break.

(The GCC example I gave above is exactly one file in one arch backend, so a
trivial patch to fix).

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/15794] move AT_* constants definition from elf.h into sysdeps (bits) directory

[carlos at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=15794

--- Comment #9 from Carlos O'Donell <carlos at redhat dot com> ---
(In reply to Adam Conrad from comment #7)
> The problem with this patch is that any code that previously included elf.h
> to get the AT_ defines is now broken (*cough*GCC*cough*).  Are you expecting
> that any such code should have an include <sys/auxv.h>, or should elf.h be
> pulling in bits/auxv.h for backward compat?

In my opinion the best way forward is:

(a) Move AT_NULL to AT_ENTRY into sysdeps/generic/bits/auvx-base.h, and
document that these 10 entries are os-independent.

(b) Move all other AT_* entries into sysdeps/unix/sysv/linux/bits/auxv.h. Have
bits/auxv.h include bits/auxv-base.h first. Other OSs can override with their
own bits/auxv.h.

(c) Have elf.h include bits/auxv-base.h (not bits/auxv.h) to provide the
os-independent auxv constants. This should prevent some source breakage and
allow the continued use of AT_NULL to AT_ENTRY Which appear to be the only set
of truly os-independent AT_* constants.

(d) Have sys/auxv.h include bits/auxv.h to make all AT_* constants available
for use with getauxval().

(e) Post those patches to libc-alpha for review, and make sure that you have
gcc, binutils, and gdb patches ready if the changes for (a), (b), and (c) break
any of the core toolchain components.

Once you get consensus from glibc it's easier to go to the other projects and
post patches. If a project uses a constant outside of the general
os-independent range (values 0-10) then you'll need a configure check to
determine if you need to include sys/auxv.h or not. It's probably sufficient to
detect if auxv.h is present, and if it is include it after including elf.h. So
you can crib up a configure check for that and use it for all projects that
need a fix (some of which might already have such a check).

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/15794] move AT_* constants definition from elf.h into sysdeps (bits) directory

[adconrad at 0c3 dot net]

https://sourceware.org/bugzilla/show_bug.cgi?id=15794

--- Comment #10 from Adam Conrad <adconrad at 0c3 dot net> ---
(In reply to Carlos O'Donell from comment #9)
> (c) Have elf.h include bits/auxv-base.h (not bits/auxv.h) to provide the
> os-independent auxv constants. This should prevent some source breakage and
> allow the continued use of AT_NULL to AT_ENTRY Which appear to be the only
> set of truly os-independent AT_* constants.

This would fix the one gcc issue I found with elf.h no longer including these
definitions.  I can scrub binutils, but given that it defines a lot of these
bits independently anyway, I imagine it'll be in good shape too.

Random userspace packages might explode here and there, but these will all be
build failures, not runtime, and we can easily weed them out in an archive
rebuild test in Debian or Ubuntu and push patches upstream for the rather tiny
set of sources that I expect will be affected by suddenly needing sys/auxv.h.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/15794] move AT_* constants definition from elf.h into sysdeps (bits) directory

[bugdal at aerifal dot cx]

https://sourceware.org/bugzilla/show_bug.cgi?id=15794

Rich Felker <bugdal at aerifal dot cx> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |bugdal at aerifal dot cx

--- Comment #12 from Rich Felker <bugdal at aerifal dot cx> ---
I think elf.h should continue to expose these constants, possibly by including
sys/auxv.h or similar. There's no sense in causing gratuitous breakage.

If incompatible header changes that break applications that are currently using
the headers *correctly* are acceptable in glibc, then certainly changes that
only break bogus applications should be acceptable; for starters, the implicit
inclusion of sys/sysmacros.h, which defines macros named major() and minor(),
from nearly every standard header, should be removed.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/15794] move AT_* constants definition from elf.h into sysdeps (bits) directory

[fweimer at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=15794

Florian Weimer <fweimer at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
              Flags|                            |security-

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/15794] move AT_* constants definition from elf.h into sysdeps (bits) directory

[cvs-commit at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=15794

--- Comment #13 from cvs-commit at gcc dot gnu.org <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Joseph Myers <jsm28@sourceware.org>:

https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=8754a4133e154ca853e6765a3fe5c7a904c77626

commit 8754a4133e154ca853e6765a3fe5c7a904c77626
Author: Joseph Myers <joseph@codesourcery.com>
Date:   Fri May 26 15:03:31 2023 +0000

    Add AT_RSEQ_* from Linux 6.3 to elf.h

    Linux 6.3 adds constants AT_RSEQ_FEATURE_SIZE and AT_RSEQ_ALIGN; add
    them to glibc's elf.h.  (Recall that, although elf.h is a
    system-independent header, so far we've put AT_* constants there even
    if Linux-specific, as discussed in bug 15794.  So rather than making
    any attempt to fix that issue, the new constants are just added there
    alongside the existing ones.)

    Tested for x86_64.

-- 
You are receiving this mail because:
You are on the CC list for the bug.